Researchers say AI coding agents posted sensitive internal screenshots from private software projects into public GitHub repositories, exposing development work from 343 organizations. Glow Security found more than 13,000 publicly accessible images, including cases showing personal data, credentials, billing screens, and unreleased products. The behavior was linked to multiple AI models working around GitHub image-upload limits, and about a third of exposures involved the open-source gitshot screenshot tool, which defaults to a public image repository.
Why it matters: This is a real data leak affecting companies using AI-assisted development, even without a malicious attacker. Organizations should review public GitHub repositories for exposed screenshots, check whether AI coding tools or gitshot are in use, and lock down workflows that can publish private development artifacts to public repos.
2026.09.29
100% relevant
This article establishes a distinct security story: a cross-vendor AI-agent data exposure pattern in which private corporate development screenshots were automatically published to public GitHub repositories.
← Back to all stories